Unitree IPO: +460% on day 1, $50B valuation — and the "ChatGPT moment" for robots promised for 2027
🔎 A robot dog worth $50 billion
On August 19, 2026, Unitree Robotics made its debut on Shanghai's STAR Market. At the close, the stock had surged 460%, bringing the robot maker's valuation to around $50 billion. An intraday peak even neared +629%, according to ABC News.
Three days later, on August 21, ACE Robotics — a Chinese competitor co-founded by a former SenseTime employee — announced that embodied intelligence would experience its "ChatGPT moment" before the end of 2027, according to Reuters.
Two announcements, the same week, the same message: China is betting that humanoid robotics is the next technological frontier, and it plans to fund it through the stock market. One uncomfortable question remains: are these billions buying an industrial revolution or viral videos of robots doing the moonwalk?
The Essentials
- Unitree Robotics raised 6.1 billion yuan ($904M) through its IPO on Shanghai's STAR Market, with a record oversubscription of 8,000 times, according to Bloomberg.
- The stock surged 460% at the close (intraday peak at +629%), valuing the company at around $50-66B depending on the source; founder Wang Xingxing (36 years old) sees his paper fortune exceed $11B, reports finance.biggo.
- ACE Robotics promises a "ChatGPT moment" for robotics by late 2027, thanks to 4-billion-parameter world models that would surpass Nvidia Cosmos 3 on certain benchmarks, according to Reuters.
- On the US side, Figure, Agility Robotics and 1X Technologies remain private, and the FCC now bans foreign robots on US soil — a total geopolitical contrast.
Recommended Tools
| Tool | Main Use | Price (August 2026, check on unitree.com) | Ideal for |
|---|---|---|---|
| Unitree G1 | Consumer humanoid robot | ~$2,700 (2024 price, Nikkei Asia) | Physical robotics development and research |
| Unitree H1 | High-performance humanoid robot | Quote on request (August 2026) | Industrial applications and logistics |
| Claude Opus 4.7 (Adaptive) (Anthropic) | Agentic model for robotic planning | Pro/Enterprise subscription (anthropic.com) | Multi-agent orchestration for robots |
| GPT-5.5 (OpenAI) | Best agentic LLM (score 98.2) | Plus/Pro subscription (openai.com) | Robot "brain", chain-of-thought reasoning |
The Unitree IPO: figures, records, and context
Unitree's stock market listing is no ordinary event, even by the frenetic standards of the STAR Market.
6.1 billion CNY raised, 8,000x oversubscription
Unitree sold its shares at an offering price set around 150 yuan. Demand exceeded supply by a factor of 8,000 — an absolute record for the STAR Market according to data compiled by AI Business.
The company thus raised 6.1 billion yuan, or approximately $904 million, placing it among the largest Chinese tech IPOs of 2026.
+460% at close: normal or delirious?
The stock opened at 1,100 yuan (+629%) before stabilizing at 845 yuan at close (+460%), reports Yahoo Finance.
These figures are spectacular, but not entirely absurd in the Chinese context. According to The Information, the Chinese A-share market shows an average first-day jump of 225% for recent IPOs. Unitree therefore more than doubles this average, without however creating an absolute precedent.
The resulting valuation fluctuates between $50 and $66 billion depending on the calculation methods (South China Morning Post cites $66 billion; other sources put it at ~$50 billion). In any case, it is more than the market capitalization of many American robotics players combined.
Wang Xingxing: 36 years old and $11 billion on paper
The founder, Daily Sabah describes him as an engineer based in Hangzhou, who launched Unitree in 2016. The company was initially known for its cheap quadruped robots — the famous "robot dogs" that flooded social media.
Today, his paper fortune exceeds 100 billion yuan (~$14 billion) according to finance.biggo. A symbol of the new generation of Chinese tech entrepreneurs who go straight from the lab to the billions.
DeepSeek and Tencent among the investors
A detail often overlooked: DeepSeek, the AI startup that shook the LLM market in late 2024, is among Unitree's investors to the tune of around $20.8M. Tencent is also a shareholder.
This convergence between software AI (DeepSeek) and physical AI (Unitree) is not insignificant. It outlines the complete ecosystem that China is trying to build: from the meilleurs LLM gratuits to embedded brains, by way of robotic bodies.
Unitree in numbers: what is this company actually selling?
Stock market enthusiasm must be put into perspective with actual revenues.
23,700 robots sold in 2024
According to Nikkei Asia, Unitree sold 23,700 units in 2024, at an average price of $2,700. This yields revenue of approximately $64 million.
Relative to a $50B valuation, the revenue/sales multiple exceeds 780x. For comparison, Nvidia at the peak of the AI bubble rarely exceeded 100x. Even the hottest pre-IPO startups generally trade between 30x and 100x.
From quadruped robots to humanoids: the transition
Unitree started with quadruped robots (Go1, Go2, A1) sold at very aggressive prices — often under $3,000, whereas Boston Dynamics charged $75,000 for its Spot.
The company has since pivoted to humanoids with the G1 and the H1. As an example, Figure Helix-02 : des robots humanoïdes accomplissent des shifts complets de 8 heures en usine — sans intervention humaine illustrates the level of autonomy the sector is aiming for. But most viral videos still show choreographed demonstrations, not autonomous productive tasks.
The promised "ChatGPT moment" by late 2027
On August 21, two days after Unitree's IPO, ACE Robotics fueled the narrative with a stunning announcement.
What ACE Robotics said
The CEO of ACE Robotics told Reuters that he expected embodied intelligence to reach its "ChatGPT moment" by late 2027.
This moment would be driven by two factors: world models capable of predicting physics and the consequences of actions, and the massive collection of real environmental data.
4 Bn parameters that beat Nvidia Cosmos 3
The most surprising claim: ACE Robotics claims that its world model of only 4 billion parameters outperforms Nvidia Cosmos 3 on certain benchmarks. If these results are independently confirmed, it would challenge the dogma that more parameters equals better physical understanding.
Nvidia launched Cosmos 3 and Isaac GR00T precisely to dominate this software layer of robotics. A 4 Bn parameter Chinese challenger beating them on their own benchmarks would be a strong signal — or a poorly designed benchmark.
Tens of millions of hours of production data
ACE Robotics claims to have deployed lightweight sensors on industrial production lines, collecting "tens of millions of hours" of physical data. This is the exact parallel to what propelled LLMs: the quantity of quality data.
The argument is solid in theory. Language models exploded thanks to terabytes of text. World models need terabytes of physical interactions. Whoever accumulates this data the fastest wins.
Is the ChatGPT parallel relevant?
The parallel has its limits. ChatGPT exploded because anyone with a browser could test an LLM. Embodied robotics requires hardware, physical space, and a cost per interaction infinitely higher than a text prompt.
The robots' "ChatGPT moment" will likely be more gradual and less democratic. But the direction is right: when a robot can learn a new task in a few hours of observation instead of months of programming, the shift will be real.
China vs United States: two opposing visions of robotics
Unitree's IPO and ACE Robotics' statements take on their full meaning in the current geopolitical context.
China opens up capital, the US closes its doors
Unitree raises $904M on the stock market, with a valuation that explodes. China is actively encouraging its robotics champions to access public markets, providing a massive financing channel.
On the other side of the Pacific Ocean, Figure AI, Agility Robotics, and 1X Technologies remain private companies, dependent on venture capital. And above all, the US FCC recently banned the import and use of foreign robots on US soil — a measure clearly targeted at Chinese manufacturers.
The hardware gap is real
In volume, Unitree dominates. Its 23,700 units in 2024 place the company far ahead of its American competitors in terms of deployment. The unit price of $2,700 makes the robot accessible to university laboratories around the world, creating a network effect.
American players like Figure (partner of OpenAI) or Agility (Digit, deployed at Amazon) are betting on quality and industrial partnerships, but are producing at much lower scales.
The "brain gap": the real battle
As finance.biggo points out, the "brain gap" is becoming the real issue. Unitree builds the bodies. But the brain — the model that drives the robot in real time — remains largely dominated by software AI players.
This is where agentic LLMs come into play. A model like GPT-5.5, which tops the agentic rankings with a score of 98.2, or Claude Opus 4.7 (94.3), can theoretically serve as a high-level planner for a robot. But latency, reliability, and physical understanding remain major obstacles.
The AI trends of recent months show a clear shift toward agentic architectures and reasoning, two critical capabilities for robotics.
Agentic LLMs: the missing link for robots
A robot without a good language model is an automaton. A robot with an agentic LLM is an autonomous agent capable of planning, adapting, and communicating.
Which models for which robotic use cases?
The current agentic ranking provides clear indications for "robot brain" candidates:
- GPT-5.5 (OpenAI) — 98.2: the most powerful, ideal for complex multi-step planning
- Gemini 3 Pro Deep Think (Google) — 95.4: excellent for spatial and physical reasoning
- Claude Opus 4.7 (Adaptive) (Anthropic) — 94.3: robustness and safety, critical when a robot operates near humans
- Kimi K2.6 (Moonshot AI) — 88.1 (self-host): interesting because it can be deployed locally, an asset for Chinese robots that cannot rely on US APIs
The Claude vs ChatGPT comparison takes on a new dimension in this context: it is no longer just a question of text quality, but of reliability for controlling physical machines. Similarly, the ChatGPT vs Gemini face-off shifts to the terrain of spatial reasoning.
The challenge of edge deployment
A robot cannot depend on a permanent cloud connection. Latency kills. An LLM must be able to run locally (edge) under strict power constraints.
This is why models like Kimi K2.6 in its self-host version or GLM-5 (Reasoning) from Z.AI are becoming strategic for the Chinese ecosystem. They offer a performance/local deployment trade-off that US models, designed for the cloud, do not always guarantee.
The meilleurs outils IA in the robotics category are increasingly integrating this edge-to-cloud dimension.
$50B Valuation: Bubble or Rational Bet?
This is the question every serious investor is asking. Here are the two camps.
The "Bubble" Argument
23,700 robots at $2,700 generate ~$64M in revenue. Valuing that at $50B implies a 780x multiple. Even assuming exponential growth, it would take years of triple-digit growth to justify this price.
Most of Unitree's demonstrations remain choreographed: robots dancing, doing backflips, walking on varied terrain. It is technically impressive, but the distance between a backflip and a robot autonomously folding laundry remains immense.
The Information rightly points out that the first days of trading on the Chinese A-share market are structurally inflated. The average of +225% is not a sign of health, but an artifact of market mechanisms (limited supply, retail speculation).
The "Rational Bet" Argument
Humanoid robotics is a potential market of several trillion dollars if it reaches the mass production stage. The global volume leader, in the era when this market explodes, deserves a significant premium.
Unitree has proven its ability to industrialize and drastically reduce costs — going from $75,000 (Spot) to $2,700 (Go2) in a few years is a real hardware accomplishment. If the same curve applies to humanoids, the market opens up massively.
Finally, the money raised ($904M) gives Unitree the means to invest heavily in the "brain gap". With DeepSeek as an investor, access to state-of-the-art models is guaranteed.
Our Take
The current price includes a huge anticipation premium. It is not a bubble in the sense of a company without revenue, but it is a price that assumes everything goes well — and in robotics, everything never goes as fast as hoped.
❌ Common mistakes
Mistake 1: Confusing a demo with productive deployment
A robot doing a backflip on stage is not a robot working 8 hours a day in a factory. Viral videos measure virality, not viability. Solution: look at real metrics for MTBF (mean time between failures), TCO (total cost of ownership), and the completion rate of unchoreographed tasks.
Mistake 2: Directly comparing the valuation multiple with SaaS companies
A SaaS startup with $64M in revenue at 780x would be absurd. But a hardware company dominating a multi-trillion-dollar emerging market plays by different rules. This doesn't mean the price is justified — it just means the comparison is inappropriate.
Mistake 3: Ignoring the geopolitical factor
Analyzing Unitree without mentioning the FCC, chip export restrictions, and China's national strategy is missing the point. Unitree's valuation isn't purely economic — it's also political.
Mistake 4: Taking the "ChatGPT moment" literally
When ACE Robotics says "late 2027", you should hear "between late 2027 and never". Forecasts in embodied AI have a historical error rate close to 100%. The directional trend is interesting; the date is just marketing.
❓ Frequently Asked Questions
How many Unitree robots have been sold?
23,700 units in 2024 at an average price of $2,700, according to Nikkei Asia. This represents approximately $64M in annual revenue.
Why did the stock surge so much on the first day?
The Chinese STAR Market structurally overreacts to IPOs (average +225%), the share offering was limited, and the "humanoid robot + AI" narrative is extremely speculative. The 8,000x oversubscription created an artificial scarcity effect.
Who are Unitree's investors?
DeepSeek (~$20.8M), Tencent, and other Chinese institutional investors. Founder Wang Xingxing retains control.
What is the "ChatGPT moment" of robotics?
It is the inflection point where a robot becomes capable of learning and executing a new physical task after a simple demonstration or natural language instruction, without reprogramming. ACE Robotics promises this by the end of 2027.
Can an LLM really pilot a robot?
Partially. Agentic LLMs like GPT-5.5 or Claude Opus 4.7 excel at high-level planning ("go get the red one in the kitchen"). But low-level motor control (balance, grasping) requires dedicated controllers, not an LLM.
Did the FCC really ban Chinese robots?
Yes. The FCC adopted rules prohibiting the import and use of robots manufactured by China-linked entities on U.S. soil, citing national security risks related to sensors and data collection.
✅ Conclusion
Unitree's IPO is a pivotal moment: the first time a humanoid robot manufacturer reaches a valuation of around $50 billion on the stock market. Whether justified or overvalued, the signal has been sent — physical robotics is now a premier financial asset, not a lab project. The Chinese stock market is financing the transition from cheap hardware to the embodied brain, with an ambitious target: late 2027 for the real tipping point. The coming months will tell if the money follows the physics or if the physics catches up with the money. Follow the tendances IA so you don't miss the next episode.
```